In this post, I’m going to talk about my journey from design leadership to building and launching my own products using agentic coding.

Transparency and disclosure note

This article contains affiliate links for Zero to Shipped, Codefa.st, Shipfa.st, and Lovable. If you make a purchase through any of these links, I’ll receive a small commission.

Starting out

Agentic coding wasn’t something I planned. It wasn’t even something I knew the name for when I started. I just wanted to build things. On my own. Without asking someone else to pick up the technical side. For years that had been my ultimate dream.

My background is in design leadership. I’ve led teams, built design capabilities and shaped end-to-end experiences across different organisations, products and services. I’ve worked with some brilliant engineers and product folks. But I’d never written much code. I was always adjacent to the making, rarely in it.

That started to shift last year. I came across Lovable and Replit, and thought, why not? What have I got to lose? I wasn’t aiming to become a developer, I desperately wanted to close the gap between my ideas and what I could bring to life.

I forced myself to focus on building confidence rather than worrying I wouldn’t achieve anything.

I first started with Replit. It gave me a window into building and the power it had.

The early days were clunky. The first go lasted about 45 minutes, I closed the lid on the MacBook and thought it was never going to be. I was going to be forever stuck making HTML/CSS prototypes…

Never give up

But I persisted, I spent hours reading, copying, pasting and debugging. I’d build something small, then bin it the next day. I’d make a little progress, hit a wall and scrap the whole thing. But every time I started over, I learned a little more. Not just about syntax or structure, but about how I worked.

The real shift came when I switched to using Cursor. Everything clicked. It felt more ‘Gav friendly’ and I felt like it knew my flow a bit better (it didn’t, obviously) but it felt smoother. The guidance felt sharper. It didn’t feel like fighting with a tool anymore, for now anyway.

Cursor’s biggest strength for me was its agent mode, which then became its multi agent + thinking mode. When I switched it on, it started acting less like a code editor and more like a pair of hands that understood what I was trying to make. It could scan my files, interpret what I was attempting and offer suggestions without needing constant hand-holding. It felt like I was sitting next to Lew, without asking him to help me.

The different agent options added another level, sometimes one agent might get caught up in itself, start hallucinating or nuke my database (yes that happened). I learned that I could switch between different assistant models, depending on what I needed. Claude Sonnet was particularly impressive for a while, Gemini was good too. Each variation offered a slightly different lens, some were better at reasoning, others at speeding up repetitive refactors or breaking down complex problems.

Sometimes I just needed help breaking a problem down. Other times I wanted full suggestions across multiple files.

I built a thing or two

I’ve now launched two projects on my own. No dev team. No agency. Just me, a few tools and a lot of time figuring things out. They weren’t perfect. That wasn’t the point. The point was that I made them real, I shipped them, I learned from them and built up more confidence in a few short months than I’d done over years.

The first was bestaithings.com, a curated directory of AI tools and products. I kept seeing the same question over and over: “What’s the best AI tool for X?” I wanted to create a clean, fast and useful site that gave people good answers. It’s simple and searchable. I used it to practice working with external APIs, implementing real-time search and handling a lightweight CMS flow. It’s the kind of project that could’ve taken weeks or more. I did it in a few focused sessions.

The second is yourseasonguide.com. After speaking with my better half and me uttering the words “I could make an app for that”, it became a reality. I pushed myself further with this project than any other that came before it. And that was the point, keep pushing, keep learning. An app that can help you choose which colours suit you the best with recommendations to go with it. I wanted to create something that felt helpful, friendly and grounded. I learned a large amount about using the correct stack, workflow and integrations. Again, it’s not a massive product. But it’s real, live and useful.

The third thing is currently in the works and will help small to medium businesses with recruitment. More to come on that in the next few weeks.

The accelerators

Let’s talk about choosing the right stack to build projects for a second because I think this may be the area that tripped me up the most. I recognised that in each of the things that made no progress, I hit a wall around the same point. I’d start with one set of tools, get lost in configuration or limitations and ditch the whole thing. I didn’t know what I should be using, how pieces fit together or which tools were just too heavy for what I needed.

Then I came across Marc Lou and Kitze, who quite honestly, has been a bit of inspiration. He has a suite of different products to help and enable builders or those wanting to build more easily. These resources made a huge difference and will make a difference depending on which stage you’re at.

Codefa.st is designed for those with entrepreneurial, it’s an accelerated learning path to build your SaaS or online business even if you have no coding experience. It teaches the essentials and helps you launch fast using AI to assist with development. Its three-part structure gives you everything from the basics to launching a product. The frontend, backend and business stack are all handled with just enough support and flexibility.

ShipFast is a Next.js boilerplate built for speed. It gives you authentication, payments, email integration and SEO setup out of the box. You don’t have to spend hours wiring things up just to see if an idea has legs. You can build the core, test it, get feedback and iterate.

Zero to Shipped is an interactive video course that walks you through the full-stack dev journey. It introduces a practical tech stack—Next.js, Blitz.js, Prisma, Mantine UI, TypeScript and Zod. It’s built for people like me who don’t need more complexity. Just a clear path from idea to working product. It covers databases, authentication, hosting and more, making it easy to get going fast with the right foundation.

Gaining confidence

These tools can help you avoid the mistakes I made early on so you can move faster, more confidently. And with fewer false starts, as it can be totally demoralizing.

Confidence didn’t come from having all the answers. It came from pushing through the mess. From not giving up when I hit a wall. From learning when to ask better questions or just take a break and come back with a fresh head.

There’s also the bit about questioning your own sanity when you end up in a literal argument with the agent.

What I’ve realised is this: the gap between idea and execution is smaller than it’s ever been. But it still takes work. Still takes showing up and putting in the effort.

If you’re thinking about building something, here are a few things that helped me get started:

• Just start. Put fear to one side, and just give it a go.
• Start small. Don’t try to build the whole thing at once. Build the part that proves the idea.
• Use tools that reduce friction. For me, that was Cursor. For you it might be something else. Just don’t let setup be the thing that stops you.
• Expect to throw work away. That’s part of the process. Every dead end teaches you something. The next thing will be better than the last thing.
• Share early. Even if it’s rough. Especially if it’s rough. Feedback helps.

And if you want something to guide you:

• CodeFast gives you the structure and support to learn and launch in a focused, practical way.
• ShipFast helps you get live without overthinking infrastructure.
• Zero to Shipped is an interactive video course that walks you through the full-stack dev journey.

This isn’t about becoming a dev. It’s about owning more of the process. About removing blockers. About feeling the momentum of making something real.

Agentic coding gave me that. And it’s reshaped how I think about ideas, execution and the creative process.

If you’re on the edge of trying it, step in. You won’t regret it.

And to all my dev friends

I know what you’ll be saying, I get it, nothing will beat a high-level professional software engineer. There’s a business proposition in there.

But please don’t underestimate the learning, development, confidence building and growth someone can go through from doing this. It might be one of the best things that has ever happened to bring more professions together.